1Locale::Maketext::TPJ13(P3eprml)Programmers Reference GLuoicdaele::Maketext::TPJ13(3pm)
2
6 Locale::Maketext::TPJ13 -- article about software localization
7
9 # This an article, not a module.
10
12 The following article by Sean M. Burke and Jordan Lachler first
13 appeared in The Perl Journal #13 and is copyright 1999 The Perl
14 Journal. It appears courtesy of Jon Orwant and The Perl Journal. This
15 document may be distributed under the same terms as Perl itself.
16
18 by Sean M. Burke and Jordan Lachler
19 This article points out cases where gettext (a common system for
21 localizing software interfaces -- i.e., making them work in the user's
22 language of choice) fails because of basic differences between human
23 languages. This article then describes Maketext, a new system capable
24 of correctly treating these differences.
25 A Localization Horror Story: It Could Happen To You
27 "There are a number of languages spoken by human beings in this
28 world."
29 -- Harald Tveit Alvestrand, in RFC 1766, "Tags for the
31 Identification of Languages"
32 Imagine that your task for the day is to localize a piece of software
34 -- and luckily for you, the only output the program emits is two
35 messages, like this:
36 I scanned 12 directories.
38 Your query matched 10 files in 4 directories.
40 So how hard could that be? You look at the code that produces the
42 first item, and it reads:
43 printf("I scanned %g directories.",
45 $directory_count);
46 You think about that, and realize that it doesn't even work right for
48 English, as it can produce this output:
49 I scanned 1 directories.
51 So you rewrite it to read:
53 printf("I scanned %g %s.",
55 $directory_count,
56 $directory_count == 1 ?
57 "directory" : "directories",
58 );
59 ...which does the Right Thing. (In case you don't recall, "%g" is for
61 locale-specific number interpolation, and "%s" is for string
62 interpolation.)
63 But you still have to localize it for all the languages you're
65 producing this software for, so you pull Locale::gettext off of CPAN so
66 you can access the "gettext" C functions you've heard are standard for
67 localization tasks.
68 And you write:
70 printf(gettext("I scanned %g %s."),
72 $dir_scan_count,
73 $dir_scan_count == 1 ?
74 gettext("directory") : gettext("directories"),
75 );
76 But you then read in the gettext manual (Drepper, Miller, and Pinard
78 1995) that this is not a good idea, since how a single word like
79 "directory" or "directories" is translated may depend on context -- and
80 this is true, since in a case language like German or Russian, you'd
81 may need these words with a different case ending in the first instance
82 (where the word is the object of a verb) than in the second instance,
83 which you haven't even gotten to yet (where the word is the object of a
84 preposition, "in %g directories") -- assuming these keep the same
85 syntax when translated into those languages.
86 So, on the advice of the gettext manual, you rewrite:
88 printf( $dir_scan_count == 1 ?
90 gettext("I scanned %g directory.") :
91 gettext("I scanned %g directories."),
92 $dir_scan_count );
93 So, you email your various translators (the boss decides that the
95 languages du jour are Chinese, Arabic, Russian, and Italian, so you
96 have one translator for each), asking for translations for "I scanned
97 %g directory." and "I scanned %g directories.". When they reply,
98 you'll put that in the lexicons for gettext to use when it localizes
99 your software, so that when the user is running under the "zh"
100 (Chinese) locale, gettext("I scanned %g directory.") will return the
101 appropriate Chinese text, with a "%g" in there where printf can then
102 interpolate $dir_scan.
103 Your Chinese translator emails right back -- he says both of these
105 phrases translate to the same thing in Chinese, because, in linguistic
106 jargon, Chinese "doesn't have number as a grammatical category" --
107 whereas English does. That is, English has grammatical rules that
108 refer to "number", i.e., whether something is grammatically singular or
109 plural; and one of these rules is the one that forces nouns to take a
110 plural suffix (generally "s") when in a plural context, as they are
111 when they follow a number other than "one" (including, oddly enough,
112 "zero"). Chinese has no such rules, and so has just the one phrase
113 where English has two. But, no problem, you can have this one Chinese
114 phrase appear as the translation for the two English phrases in the
115 "zh" gettext lexicon for your program.
116 Emboldened by this, you dive into the second phrase that your software
118 needs to output: "Your query matched 10 files in 4 directories.". You
119 notice that if you want to treat phrases as indivisible, as the gettext
120 manual wisely advises, you need four cases now, instead of two, to
121 cover the permutations of singular and plural on the two items,
122 $dir_count and $file_count. So you try this:
123 printf( $file_count == 1 ?
125 ( $directory_count == 1 ?
126 gettext("Your query matched %g file in %g directory.") :
127 gettext("Your query matched %g file in %g directories.") ) :
128 ( $directory_count == 1 ?
129 gettext("Your query matched %g files in %g directory.") :
130 gettext("Your query matched %g files in %g directories.") ),
131 $file_count, $directory_count,
132 );
133 (The case of "1 file in 2 [or more] directories" could, I suppose,
135 occur in the case of symlinking or something of the sort.)
136 It occurs to you that this is not the prettiest code you've ever
138 written, but this seems the way to go. You mail off to the translators
139 asking for translations for these four cases. The Chinese guy replies
140 with the one phrase that these all translate to in Chinese, and that
141 phrase has two "%g"s in it, as it should -- but there's a problem. He
142 translates it word-for-word back: "In %g directories contains %g files
143 match your query." The %g slots are in an order reverse to what they
144 are in English. You wonder how you'll get gettext to handle that.
145 But you put it aside for the moment, and optimistically hope that the
147 other translators won't have this problem, and that their languages
148 will be better behaved -- i.e., that they will be just like English.
149 But the Arabic translator is the next to write back. First off, your
151 code for "I scanned %g directory." or "I scanned %g directories."
152 assumes there's only singular or plural. But, to use linguistic jargon
153 again, Arabic has grammatical number, like English (but unlike
154 Chinese), but it's a three-term category: singular, dual, and plural.
155 In other words, the way you say "directory" depends on whether there's
156 one directory, or two of them, or more than two of them. Your test of
157 "($directory == 1)" no longer does the job. And it means that where
158 English's grammatical category of number necessitates only the two
159 permutations of the first sentence based on "directory [singular]" and
160 "directories [plural]", Arabic has three -- and, worse, in the second
161 sentence ("Your query matched %g file in %g directory."), where English
162 has four, Arabic has nine. You sense an unwelcome, exponential trend
163 taking shape.
164 Your Italian translator emails you back and says that "I searched 0
166 directories" (a possible English output of your program) is stilted,
167 and if you think that's fine English, that's your problem, but that
168 just will not do in the language of Dante. He insists that where
169 $directory_count is 0, your program should produce the Italian text for
170 "I didn't scan any directories.". And ditto for "I didn't match any
171 files in any directories", although he says the last part about "in any
172 directories" should probably just be left off.
173 You wonder how you'll get gettext to handle this; to accomodate the
175 ways Arabic, Chinese, and Italian deal with numbers in just these few
176 very simple phrases, you need to write code that will ask gettext for
177 different queries depending on whether the numerical values in question
178 are 1, 2, more than 2, or in some cases 0, and you still haven't
179 figured out the problem with the different word order in Chinese.
180 Then your Russian translator calls on the phone, to personally tell you
182 the bad news about how really unpleasant your life is about to become:
183 Russian, like German or Latin, is an inflectional language; that is,
185 nouns and adjectives have to take endings that depend on their case
186 (i.e., nominative, accusative, genitive, etc...) -- which is roughly a
187 matter of what role they have in syntax of the sentence -- as well as
188 on the grammatical gender (i.e., masculine, feminine, neuter) and
189 number (i.e., singular or plural) of the noun, as well as on the
190 declension class of the noun. But unlike with most other inflected
191 languages, putting a number-phrase (like "ten" or "forty-three", or
192 their Arabic numeral equivalents) in front of noun in Russian can
193 change the case and number that noun is, and therefore the endings you
194 have to put on it.
195 He elaborates: In "I scanned %g directories", you'd expect
197 "directories" to be in the accusative case (since it is the direct
198 object in the sentence) and the plural number, except where
199 $directory_count is 1, then you'd expect the singular, of course. Just
200 like Latin or German. But! Where $directory_count % 10 is 1 ("%" for
201 modulo, remember), assuming $directory count is an integer, and except
202 where $directory_count % 100 is 11, "directories" is forced to become
203 grammatically singular, which means it gets the ending for the
204 accusative singular... You begin to visualize the code it'd take to
205 test for the problem so far, and still work for Chinese and Arabic and
206 Italian, and how many gettext items that'd take, but he keeps going...
207 But where $directory_count % 10 is 2, 3, or 4 (except where
208 $directory_count % 100 is 12, 13, or 14), the word for "directories" is
209 forced to be genitive singular -- which means another ending... The
210 room begins to spin around you, slowly at first... But with all other
211 integer values, since "directory" is an inanimate noun, when preceded
212 by a number and in the nominative or accusative cases (as it is here,
213 just your luck!), it does stay plural, but it is forced into the
214 genitive case -- yet another ending... And you never hear him get to
215 the part about how you're going to run into similar (but maybe subtly
216 different) problems with other Slavic languages like Polish, because
217 the floor comes up to meet you, and you fade into unconsciousness.
218 The above cautionary tale relates how an attempt at localization can
220 lead from programmer consternation, to program obfuscation, to a need
221 for sedation. But careful evaluation shows that your choice of tools
222 merely needed further consideration.
223 The Linguistic View
225 "It is more complicated than you think."
226 -- The Eighth Networking Truth, from RFC 1925
228 The field of Linguistics has expended a great deal of effort over the
230 past century trying to find grammatical patterns which hold across
231 languages; it's been a constant process of people making
232 generalizations that should apply to all languages, only to find out
233 that, all too often, these generalizations fail -- sometimes failing
234 for just a few languages, sometimes whole classes of languages, and
235 sometimes nearly every language in the world except English. Broad
236 statistical trends are evident in what the "average language" is like
237 as far as what its rules can look like, must look like, and cannot look
238 like. But the "average language" is just as unreal a concept as the
239 "average person" -- it runs up against the fact no language (or person)
240 is, in fact, average. The wisdom of past experience leads us to
241 believe that any given language can do whatever it wants, in any order,
242 with appeal to any kind of grammatical categories wants -- case,
243 number, tense, real or metaphoric characteristics of the things that
244 words refer to, arbitrary or predictable classifications of words based
245 on what endings or prefixes they can take, degree or means of certainty
246 about the truth of statements expressed, and so on, ad infinitum.
247 Mercifully, most localization tasks are a matter of finding ways to
249 translate whole phrases, generally sentences, where the context is
250 relatively set, and where the only variation in content is usually in a
251 number being expressed -- as in the example sentences above.
252 Translating specific, fully-formed sentences is, in practice, fairly
253 foolproof -- which is good, because that's what's in the phrasebooks
254 that so many tourists rely on. Now, a given phrase (whether in a
255 phrasebook or in a gettext lexicon) in one language might have a
256 greater or lesser applicability than that phrase's translation into
257 another language -- for example, strictly speaking, in Arabic, the
258 "your" in "Your query matched..." would take a different form depending
259 on whether the user is male or female; so the Arabic translation
260 "your[feminine] query" is applicable in fewer cases than the
261 corresponding English phrase, which doesn't distinguish the user's
262 gender. (In practice, it's not feasable to have a program know the
263 user's gender, so the masculine "you" in Arabic is usually used, by
264 default.)
265 But in general, such surprises are rare when entire sentences are being
267 translated, especially when the functional context is restricted to
268 that of a computer interacting with a user either to convey a fact or
269 to prompt for a piece of information. So, for purposes of
270 localization, translation by phrase (generally by sentence) is both the
271 simplest and the least problematic.
272 Breaking gettext
274 "It Has To Work."
275 -- First Networking Truth, RFC 1925
277 Consider that sentences in a tourist phrasebook are of two types: ones
279 like "How do I get to the marketplace?" that don't have any blanks to
280 fill in, and ones like "How much do these ___ cost?", where there's one
281 or more blanks to fill in (and these are usually linked to a list of
282 words that you can put in that blank: "fish", "potatoes", "tomatoes",
283 etc.) The ones with no blanks are no problem, but the fill-in-the-
284 blank ones may not be really straightforward. If it's a Swahili
285 phrasebook, for example, the authors probably didn't bother to tell you
286 the complicated ways that the verb "cost" changes its inflectional
287 prefix depending on the noun you're putting in the blank. The trader
288 in the marketplace will still understand what you're saying if you say
289 "how much do these potatoes cost?" with the wrong inflectional prefix
290 on "cost". After all, you can't speak proper Swahili, you're just a
291 tourist. But while tourists can be stupid, computers are supposed to
292 be smart; the computer should be able to fill in the blank, and still
293 have the results be grammatical.
294 In other words, a phrasebook entry takes some values as parameters (the
296 things that you fill in the blank or blanks), and provides a value
297 based on these parameters, where the way you get that final value from
298 the given values can, properly speaking, involve an arbitrarily complex
299 series of operations. (In the case of Chinese, it'd be not at all
300 complex, at least in cases like the examples at the beginning of this
301 article; whereas in the case of Russian it'd be a rather complex series
302 of operations. And in some languages, the complexity could be spread
303 around differently: while the act of putting a number-expression in
304 front of a noun phrase might not be complex by itself, it may change
305 how you have to, for example, inflect a verb elsewhere in the sentence.
306 This is what in syntax is called "long-distance dependencies".)
307 This talk of parameters and arbitrary complexity is just another way to
309 say that an entry in a phrasebook is what in a programming language
310 would be called a "function". Just so you don't miss it, this is the
311 crux of this article: A phrase is a function; a phrasebook is a bunch
312 of functions.
313 The reason that using gettext runs into walls (as in the above second-
315 person horror story) is that you're trying to use a string (or worse, a
316 choice among a bunch of strings) to do what you really need a function
317 for -- which is futile. Preforming (s)printf interpolation on the
318 strings which you get back from gettext does allow you to do some
319 common things passably well... sometimes... sort of; but, to paraphrase
320 what some people say about "csh" script programming, "it fools you into
321 thinking you can use it for real things, but you can't, and you don't
322 discover this until you've already spent too much time trying, and by
323 then it's too late."
324 Replacing gettext
326 So, what needs to replace gettext is a system that supports lexicons of
327 functions instead of lexicons of strings. An entry in a lexicon from
328 such a system should not look like this:
329 "J'ai trouv\xE9 %g fichiers dans %g r\xE9pertoires"
331 [\xE9 is e-acute in Latin-1. Some pod renderers would scream if I used
333 the actual character here. -- SB]
334 but instead like this, bearing in mind that this is just a first stab:
336 sub I_found_X1_files_in_X2_directories {
338 my( $files, $dirs ) = @_[0,1];
339 $files = sprintf("%g %s", $files,
340 $files == 1 ? 'fichier' : 'fichiers');
341 $dirs = sprintf("%g %s", $dirs,
342 $dirs == 1 ? "r\xE9pertoire" : "r\xE9pertoires");
343 return "J'ai trouv\xE9 $files dans $dirs.";
344 }
345 Now, there's no particularly obvious way to store anything but strings
347 in a gettext lexicon; so it looks like we just have to start over and
348 make something better, from scratch. I call my shot at a gettext-
349 replacement system "Maketext", or, in CPAN terms, Locale::Maketext.
350 When designing Maketext, I chose to plan its main features in terms of
352 "buzzword compliance". And here are the buzzwords:
353 Buzzwords: Abstraction and Encapsulation
355 The complexity of the language you're trying to output a phrase in is
356 entirely abstracted inside (and encapsulated within) the Maketext
357 module for that interface. When you call:
358 print $lang->maketext("You have [quant,_1,piece] of new mail.",
360 scalar(@messages));
361 you don't know (and in fact can't easily find out) whether this will
363 involve lots of figuring, as in Russian (if $lang is a handle to the
364 Russian module), or relatively little, as in Chinese. That kind of
365 abstraction and encapsulation may encourage other pleasant buzzwords
366 like modularization and stratification, depending on what design
367 decisions you make.
368 Buzzword: Isomorphism
370 "Isomorphism" means "having the same structure or form"; in discussions
371 of program design, the word takes on the special, specific meaning that
372 your implementation of a solution to a problem has the same structure
373 as, say, an informal verbal description of the solution, or maybe of
374 the problem itself. Isomorphism is, all things considered, a good
375 thing -- it's what problem-solving (and solution-implementing) should
376 look like.
377 What's wrong the with gettext-using code like this...
379 printf( $file_count == 1 ?
381 ( $directory_count == 1 ?
382 "Your query matched %g file in %g directory." :
383 "Your query matched %g file in %g directories." ) :
384 ( $directory_count == 1 ?
385 "Your query matched %g files in %g directory." :
386 "Your query matched %g files in %g directories." ),
387 $file_count, $directory_count,
388 );
389 is first off that it's not well abstracted -- these ways of testing for
391 grammatical number (as in the expressions like "foo == 1 ?
392 singular_form : plural_form") should be abstracted to each language
393 module, since how you get grammatical number is language-specific.
394 But second off, it's not isomorphic -- the "solution" (i.e., the
396 phrasebook entries) for Chinese maps from these four English phrases to
397 the one Chinese phrase that fits for all of them. In other words, the
398 informal solution would be "The way to say what you want in Chinese is
399 with the one phrase 'For your question, in Y directories you would find
400 X files'" -- and so the implemented solution should be, isomorphically,
401 just a straightforward way to spit out that one phrase, with numerals
402 properly interpolated. It shouldn't have to map from the complexity of
403 other languages to the simplicity of this one.
404 Buzzword: Inheritance
406 There's a great deal of reuse possible for sharing of phrases between
407 modules for related dialects, or for sharing of auxiliary functions
408 between related languages. (By "auxiliary functions", I mean functions
409 that don't produce phrase-text, but which, say, return an answer to
410 "does this number require a plural noun after it?". Such auxiliary
411 functions would be used in the internal logic of functions that
412 actually do produce phrase-text.)
413 In the case of sharing phrases, consider that you have an interface
415 already localized for American English (probably by having been written
416 with that as the native locale, but that's incidental). Localizing it
417 for UK English should, in practical terms, be just a matter of running
418 it past a British person with the instructions to indicate what few
419 phrases would benefit from a change in spelling or possibly minor
420 rewording. In that case, you should be able to put in the UK English
421 localization module only those phrases that are UK-specific, and for
422 all the rest, inherit from the American English module. (And I expect
423 this same situation would apply with Brazilian and Continental
424 Portugese, possbily with some very closely related languages like Czech
425 and Slovak, and possibly with the slightly different "versions" of
426 written Mandarin Chinese, as I hear exist in Taiwan and mainland
427 China.)
428 As to sharing of auxiliary functions, consider the problem of Russian
430 numbers from the beginning of this article; obviously, you'd want to
431 write only once the hairy code that, given a numeric value, would
432 return some specification of which case and number a given quanitified
433 noun should use. But suppose that you discover, while localizing an
434 interface for, say, Ukranian (a Slavic language related to Russian,
435 spoken by several million people, many of whom would be relieved to
436 find that your Web site's or software's interface is available in their
437 language), that the rules in Ukranian are the same as in Russian for
438 quantification, and probably for many other grammatical functions.
439 While there may well be no phrases in common between Russian and
440 Ukranian, you could still choose to have the Ukranian module inherit
441 from the Russian module, just for the sake of inheriting all the
442 various grammatical methods. Or, probably better organizationally, you
443 could move those functions to a module called "_E_Slavic" or something,
444 which Russian and Ukranian could inherit useful functions from, but
445 which would (presumably) provide no lexicon.
446 Buzzword: Concision
448 Okay, concision isn't a buzzword. But it should be, so I decree that
449 as a new buzzword, "concision" means that simple common things should
450 be expressible in very few lines (or maybe even just a few characters)
451 of code -- call it a special case of "making simple things easy and
452 hard things possible", and see also the role it played in the
453 MIDI::Simple language, discussed elsewhere in this issue [TPJ#13].
454 Consider our first stab at an entry in our "phrasebook of functions":
456 sub I_found_X1_files_in_X2_directories {
458 my( $files, $dirs ) = @_[0,1];
459 $files = sprintf("%g %s", $files,
460 $files == 1 ? 'fichier' : 'fichiers');
461 $dirs = sprintf("%g %s", $dirs,
462 $dirs == 1 ? "r\xE9pertoire" : "r\xE9pertoires");
463 return "J'ai trouv\xE9 $files dans $dirs.";
464 }
465 You may sense that a lexicon (to use a non-committal catch-all term for
467 a collection of things you know how to say, regardless of whether
468 they're phrases or words) consisting of functions expressed as above
469 would make for rather long-winded and repetitive code -- even if you
470 wisely rewrote this to have quantification (as we call adding a number
471 expression to a noun phrase) be a function called like:
472 sub I_found_X1_files_in_X2_directories {
474 my( $files, $dirs ) = @_[0,1];
475 $files = quant($files, "fichier");
476 $dirs = quant($dirs, "r\xE9pertoire");
477 return "J'ai trouv\xE9 $files dans $dirs.";
478 }
479 And you may also sense that you do not want to bother your translators
481 with having to write Perl code -- you'd much rather that they spend
482 their very costly time on just translation. And this is to say nothing
483 of the near impossibility of finding a commercial translator who would
484 know even simple Perl.
485 In a first-hack implementation of Maketext, each language-module's
487 lexicon looked like this:
488 %Lexicon = (
490 "I found %g files in %g directories"
491 => sub {
492 my( $files, $dirs ) = @_[0,1];
493 $files = quant($files, "fichier");
494 $dirs = quant($dirs, "r\xE9pertoire");
495 return "J'ai trouv\xE9 $files dans $dirs.";
496 },
497 ... and so on with other phrase => sub mappings ...
498 );
499 but I immediately went looking for some more concise way to basically
501 denote the same phrase-function -- a way that would also serve to
502 concisely denote most phrase-functions in the lexicon for most
503 languages. After much time and even some actual thought, I decided on
504 this system:
505 * Where a value in a %Lexicon hash is a contentful string instead of an
507 anonymous sub (or, conceivably, a coderef), it would be interpreted as
508 a sort of shorthand expression of what the sub does. When accessed for
509 the first time in a session, it is parsed, turned into Perl code, and
510 then eval'd into an anonymous sub; then that sub replaces the original
511 string in that lexicon. (That way, the work of parsing and evaling the
512 shorthand form for a given phrase is done no more than once per
513 session.)
514 * Calls to "maketext" (as Maketext's main function is called) happen
516 thru a "language session handle", notionally very much like an IO
517 handle, in that you open one at the start of the session, and use it
518 for "sending signals" to an object in order to have it return the text
519 you want.
520 So, this:
522 $lang->maketext("You have [quant,_1,piece] of new mail.",
524 scalar(@messages));
525 basically means this: look in the lexicon for $lang (which may inherit
527 from any number of other lexicons), and find the function that we
528 happen to associate with the string "You have [quant,_1,piece] of new
529 mail" (which is, and should be, a functioning "shorthand" for this
530 function in the native locale -- English in this case). If you find
531 such a function, call it with $lang as its first parameter (as if it
532 were a method), and then a copy of scalar(@messages) as its second, and
533 then return that value. If that function was found, but was in string
534 shorthand instead of being a fully specified function, parse it and
535 make it into a function before calling it the first time.
536 * The shorthand uses code in brackets to indicate method calls that
538 should be performed. A full explanation is not in order here, but a
539 few examples will suffice:
540 "You have [quant,_1,piece] of new mail."
542 The above code is shorthand for, and will be interpreted as, this:
544 sub {
546 my $handle = $_[0];
547 my(@params) = @_;
548 return join '',
549 "You have ",
550 $handle->quant($params[1], 'piece'),
551 "of new mail.";
552 }
553 where "quant" is the name of a method you're using to quantify the noun
555 "piece" with the number $params[0].
556 A string with no brackety calls, like this:
558 "Your search expression was malformed."
560 is somewhat of a degerate case, and just gets turned into:
562 sub { return "Your search expression was malformed." }
564 However, not everything you can write in Perl code can be written in
566 the above shorthand system -- not by a long shot. For example,
567 consider the Italian translator from the beginning of this article, who
568 wanted the Italian for "I didn't find any files" as a special case,
569 instead of "I found 0 files". That couldn't be specified (at least not
570 easily or simply) in our shorthand system, and it would have to be
571 written out in full, like this:
572 sub { # pretend the English strings are in Italian
574 my($handle, $files, $dirs) = @_[0,1,2];
575 return "I didn't find any files" unless $files;
576 return join '',
577 "I found ",
578 $handle->quant($files, 'file'),
579 " in ",
580 $handle->quant($dirs, 'directory'),
581 ".";
582 }
583 Next to a lexicon full of shorthand code, that sort of sticks out like
585 a sore thumb -- but this is a special case, after all; and at least
586 it's possible, if not as concise as usual.
587 As to how you'd implement the Russian example from the beginning of the
589 article, well, There's More Than One Way To Do It, but it could be
590 something like this (using English words for Russian, just so you know
591 what's going on):
592 "I [quant,_1,directory,accusative] scanned."
594 This shifts the burden of complexity off to the quant method. That
596 method's parameters are: the numeric value it's going to use to
597 quantify something; the Russian word it's going to quantify; and the
598 parameter "accusative", which you're using to mean that this sentence's
599 syntax wants a noun in the accusative case there, although that
600 quantification method may have to overrule, for grammatical reasons you
601 may recall from the beginning of this article.
602 Now, the Russian quant method here is responsible not only for
604 implementing the strange logic necessary for figuring out how Russian
605 number-phrases impose case and number on their noun-phrases, but also
606 for inflecting the Russian word for "directory". How that inflection
607 is to be carried out is no small issue, and among the solutions I've
608 seen, some (like variations on a simple lookup in a hash where all
609 possible forms are provided for all necessary words) are
610 straightforward but can become cumbersome when you need to inflect more
611 than a few dozen words; and other solutions (like using algorithms to
612 model the inflections, storing only root forms and irregularities) can
613 involve more overhead than is justifiable for all but the largest
614 lexicons.
615 Mercifully, this design decision becomes crucial only in the hairiest
617 of inflected languages, of which Russian is by no means the worst case
618 scenario, but is worse than most. Most languages have simpler
619 inflection systems; for example, in English or Swahili, there are
620 generally no more than two possible inflected forms for a given noun
621 ("error/errors"; "kosa/makosa"), and the rules for producing these
622 forms are fairly simple -- or at least, simple rules can be formulated
623 that work for most words, and you can then treat the exceptions as just
624 "irregular", at least relative to your ad hoc rules. A simpler
625 inflection system (simpler rules, fewer forms) means that design
626 decisions are less crucial to maintaining sanity, whereas the same
627 decisions could incur overhead-versus-scalability problems in languages
628 like Russian. It may also be likely that code (possibly in Perl, as
629 with Lingua::EN::Inflect, for English nouns) has already been written
630 for the language in question, whether simple or complex.
631 Moreover, a third possibility may even be simpler than anything
633 discussed above: "Just require that all possible (or at least
634 applicable) forms be provided in the call to the given language's quant
635 method, as in:"
636 "I found [quant,_1,file,files]."
638 That way, quant just has to chose which form it needs, without having
640 to look up or generate anything. While possibly not optimal for
641 Russian, this should work well for most other languages, where
642 quantification is not as complicated an operation.
643 The Devil in the Details
645 There's plenty more to Maketext than described above -- for example,
646 there's the details of how language tags ("en-US", "i-pwn", "fi", etc.)
647 or locale IDs ("en_US") interact with actual module naming
648 ("BogoQuery/Locale/en_us.pm"), and what magic can ensue; there's the
649 details of how to record (and possibly negotiate) what character
650 encoding Maketext will return text in (UTF8? Latin-1? KOI8?). There's
651 the interesting fact that Maketext is for localization, but nowhere
652 actually has a ""use locale;"" anywhere in it. For the curious,
653 there's the somewhat frightening details of how I actually implement
654 something like data inheritance so that searches across modules'
655 %Lexicon hashes can parallel how Perl implements method inheritance.
656 And, most importantly, there's all the practical details of how to
658 actually go about deriving from Maketext so you can use it for your
659 interfaces, and the various tools and conventions for starting out and
660 maintaining individual language modules.
661 That is all covered in the documentation for Locale::Maketext and the
663 modules that come with it, available in CPAN. After having read this
664 article, which covers the why's of Maketext, the documentation, which
665 covers the how's of it, should be quite straightfoward.
666 The Proof in the Pudding: Localizing Web Sites
668 Maketext and gettext have a notable difference: gettext is in C,
669 accessible thru C library calls, whereas Maketext is in Perl, and
670 really can't work without a Perl interpreter (although I suppose
671 something like it could be written for C). Accidents of history (and
672 not necessarily lucky ones) have made C++ the most common language for
673 the implementation of applications like word processors, Web browsers,
674 and even many in-house applications like custom query systems. Current
675 conditions make it somewhat unlikely that the next one of any of these
676 kinds of applications will be written in Perl, albeit clearly more for
677 reasons of custom and inertia than out of consideration of what is the
678 right tool for the job.
679 However, other accidents of history have made Perl a well-accepted
681 language for design of server-side programs (generally in CGI form) for
682 Web site interfaces. Localization of static pages in Web sites is
683 trivial, feasable either with simple language-negotiation features in
684 servers like Apache, or with some kind of server-side inclusions of
685 language-appropriate text into layout templates. However, I think that
686 the localization of Perl-based search systems (or other kinds of
687 dynamic content) in Web sites, be they public or access-restricted, is
688 where Maketext will see the greatest use.
689 I presume that it would be only the exceptional Web site that gets
691 localized for English and Chinese and Italian and Arabic and Russian,
692 to recall the languages from the beginning of this article -- to say
693 nothing of German, Spanish, French, Japanese, Finnish, and Hindi, to
694 name a few languages that benefit from large numbers of programmers or
695 Web viewers or both.
696 However, the ever-increasing internationalization of the Web (whether
698 measured in terms of amount of content, of numbers of content writers
699 or programmers, or of size of content audiences) makes it increasingly
700 likely that the interface to the average Web-based dynamic content
701 service will be localized for two or maybe three languages. It is my
702 hope that Maketext will make that task as simple as possible, and will
703 remove previous barriers to localization for languages dissimilar to
704 English.
705 __END__
707 Sean M. Burke (sburke@cpan.org) has a Master's in linguistics from
709 Northwestern University; he specializes in language technology. Jordan
710 Lachler (lachler@unm.edu) is a PhD student in the Department of
711 Linguistics at the University of New Mexico; he specializes in
712 morphology and pedagogy of North American native languages.
713 References
715 Alvestrand, Harald Tveit. 1995. RFC 1766: Tags for the Identification
716 of Languages. "ftp://ftp.isi.edu/in-notes/rfc1766.txt" [Now see RFC
717 3066.]
718 Callon, Ross, editor. 1996. RFC 1925: The Twelve Networking Truths.
720 "ftp://ftp.isi.edu/in-notes/rfc1925.txt"
721 Drepper, Ulrich, Peter Miller, and Francois Pinard. 1995-2001. GNU
723 "gettext". Available in "ftp://prep.ai.mit.edu/pub/gnu/", with
724 extensive docs in the distribution tarball. [Since I wrote this
725 article in 1998, I now see that the gettext docs are now trying more to
726 come to terms with plurality. Whether useful conclusions have come
727 from it is another question altogether. -- SMB, May 2001]
728 Forbes, Nevill. 1964. Russian Grammar. Third Edition, revised by J.
730 C. Dumbreck. Oxford University Press.
731perl v5.12.4 2011-06-07 Locale::Maketext::TPJ13(3pm)